Bug 11634 (DMARCREJECT) - Configure DMARC policy to p=reject
Summary: Configure DMARC policy to p=reject
Status: CLOSED FIXED
Alias: DMARCREJECT
Product: Infrastructure
Classification: Unclassified
Component: Mail & Mailing Lists (show other bugs)
Version: unspecified
Hardware: unspecified Unspecified
: - Unknown - - Unknown -
Assignee: Peter Müller
QA Contact: Peter Müller
URL:
Keywords:
Depends on: 11635 11654 11655 11676 11703 11902
Blocks:
  Show dependency treegraph
 
Reported: 2018-02-21 14:09 UTC by Michael Tremer
Modified: 2022-05-16 19:28 UTC (History)
1 user (show)

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Michael Tremer 2018-02-21 14:09:08 UTC
This is our ultimate goal which is a not possible to be done short-term. Therefore this is a tracking bug that is supposed to track all steps that are necessary for this to be achieved.
Comment 1 Peter Müller 2018-03-01 20:00:44 UTC
(In reply to Michael Tremer from comment #0)
> This is our ultimate goal which is a not possible to be done short-term.
> Therefore this is a tracking bug that is supposed to track all steps that
> are necessary for this to be achieved.
Well, we have do differ between _sending_ and _receiving_ mails.

Looking at the sender side, we have to make sure all mails can be validated (correct DKIM signature) and there are no mails with "internal" accounts - such as nginx - left. ARC is needed here, too, to add the Authentication-headers before a mail is delivered to the mailing list.

Looking at the receiving side, we need to deploy ARC and it probably will take some time until we can switch to DMARC reject mode, since mailing lists are causing a lot of trouble here at the moment.
Comment 2 Michael Tremer 2018-03-01 21:55:57 UTC
Can you create the appropriate tickets to roll out ARC?
Comment 3 Peter Müller 2018-03-03 21:37:53 UTC
(In reply to Michael Tremer from comment #2)
> Can you create the appropriate tickets to roll out ARC?
I am currently researching abount appropriate software (openarc vs. authenticating-milter), but as soon I am through this, tickets will be created.
Comment 4 Peter Müller 2019-09-01 07:53:55 UTC
Well, this is actually ON_DEV by now. As soon as we are done with new mail01, I can test ARC signing for Mailman.
Comment 5 Peter Müller 2020-04-01 12:27:30 UTC
ARC signing is now working. Two things need to be solved before we can switch to DMARC reject policy:

(a) DMARC reports from Google look bad - DKIM signatures are invalid for almost any messages. We do not observe this for any other DMARC reporter but I have no idea about the root cause.

(b) Mailman sometimes renders DKIM signatures invalid (see #11635).
Comment 6 Michael Tremer 2020-04-01 13:20:58 UTC
These are both very severe blockers.

The vast majority of emails that we sent is going to Google, unfortunately.

And we need to be able to rely on emails reaching their destinations. Of course we cannot bounce-handle a large amount of emails.
Comment 7 Peter Müller 2020-10-27 14:48:32 UTC
> (a) DMARC reports from Google look bad - DKIM signatures are invalid for
> almost any messages. We do not observe this for any other DMARC reporter but
> I have no idea about the root cause.

This turned out to be a bug in the DMARC reports parser we use, the actual results from Google look quite good. As soon as this is fixed upstream, we will be able to see valid data again.
Comment 8 Peter Müller 2021-01-16 10:23:03 UTC
The DMARC reports parsing issue has been fixed meanwhile.

For some reason, Mailman behaves better (at least the DMARC reports suggest this), which is why I propose to start by setting ipfire.org's DMARC policy to "quarantine" for 10% of the mail volume.
Comment 9 Michael Tremer 2021-01-25 19:48:24 UTC
Did we learn anything, yet?
Comment 10 Peter Müller 2021-01-26 15:55:41 UTC
> Did we learn anything, yet?

Not really, aside from me being surprised how clean our DMARC reports are. On the
other hand, the number of entities we receive reports from is not very diverse, with
GMail of course leading the list.

Sometimes, Mailman still seems to break DKIM signatures, for a reason yet to be
investigated further. Since this only happens for some - but not all - UTF-8 mails
without a Content-Type header (as emitted by "git send-email"), I am tempted to blame
Python's email module here. However, I have spent hours on trying to reproduce this
- as of today, without success. :-/

Mail.ru occasionally reports some messages spewed out by various dial-up IPs in
Russia and some countries in Eastern Europe (I am puzzled to see they even accept
mails from such IP ranges - some of them do not even have a PTR). While their number
is very low, I have not observed Mail.ru to apply a quarantine policy on them.

After the mass mailing following your blog post the other day, I have now increased
the percentage of IPFire.org's DMARC policy to 25 %.
Comment 11 Peter Müller 2021-10-26 18:09:03 UTC
> After the mass mailing following your blog post the other day, I have now increased the percentage of IPFire.org's DMARC policy to 25 %.

Meanwhile, for the records, we run ipfire.org with a DMARC quarantine policy for 100 % of the mail traffic for several months. I am unaware of any issues.